47 new T dwarfs from the UKIDSS Large Area Survey Free

Abstract

1 INTRODUCTION

The study of substellar objects presents a number of important opportunities for extending our understanding of star and planet formation, both through detailed study of individual systems and through statistical population studies. The statistical characteristics of the substellar population, such as binary fraction and distribution, and the form of the initial mass function (IMF; Salpeter 1955) across the entire substellar regime provide crucial observational constraints for models of star and planet formation, which currently offer a number of alternative formation scenarios that depend on differing balances of the dominant physics across the low-mass stellar–substellar mass spectrum (e.g. Bate & Bonnell 2005; Padoan & Nordlund 2002, and references therein). The crucial first step for any observational effort to address these issues is the initial identification of a statistically useful sample of brown dwarfs, the first of which were not discovered until the mid-1990s.

The majority of brown dwarfs have been identified via one of two routes: the mining of wide-field surveys such as the Sloan Digital Sky Survey (SDSS; York et al. 2000) and the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006) to find nearby L and T dwarf members of the field population and deep optical and near-infrared surveys of the young clusters and OB associations (e.g. Lucas & Roche 2000; Zapatero Osorio et al. 2000; Caballero et al. 2007; Bihain et al. 2009). Most recently, the United Kingdom Infrared Telescope (UKIRT) Infrared Deep Sky Survey (UKIDSS) Galactic Clusters Survey has significantly improved the substellar sample across a number of young regions (e.g. Lodieu, Hambly & Jameson 2006; Lodieu et al. 2007a, 2009).

To date, the results from clusters and associations have dominated the study of the low-mass extreme of the IMF due largely to the assumption of coevality in clusters that allows the mass–age degeneracy for substellar objects to be broken. A number of determinations have been published that are broadly in agreement across the ∼0.1–0.03 M_⊙ range (e.g. Barrado y Navascués et al. 2002; Moraux et al. 2003, 2007; Lodieu et al. 2009).

It is the age–mass degeneracy that has hampered efforts to measure the form of the substellar IMF from analysis of the local field population of L and T dwarfs. Although the determination of masses for individual field brown dwarfs is currently prevented by uncertainty about their age, the field mass function can still be constrained through comparison of the observed luminosity function or spectral type distribution with those predicted by Monte Carlo simulations for various star formation histories and underlying mass functions (e.g. Chabrier 2002; Burgasser 2004; Deacon & Hambly 2006). Allen et al. (2005) have applied a different statistical approach to solving this problem, by using Bayesian inference to evaluate the probabilities of different underlying mass functions for space densities of M, L and T dwarfs from 2MASS and SDSS, and estimated the age distribution of the field population.

Searches of the SDSS and 2MASS data sets have resulted in the discovery of over 500 L dwarfs and more than 100 T dwarfs in the field (see http://www.DwarfArchives.org for an up-to-date list of published objects). However, this sample is dominated by objects earlier than type T6, with just 26 objects identified in 2MASS and SDSS with types T6 or later and just a handful of type T8. And it is this population of objects in the ≥T6 range that is most sensitive to the underlying mass function, whilst the spectral type distribution of earlier objects depends more strongly on the Galactic formation history (see Burgasser 2004, fig. 5).

The Large Area Survey (LAS) of the UKIDSS has now successfully extended the T dwarf sample to types later than those first revealed by the 2MASS and SDSS surveys (Warren et al. 2007; Burningham et al. 2008, 2009) and is now identifying a substantial sample of late-type T dwarfs that will be ideal for constraining the substellar mass function in the field. In this paper we extend our searches of earlier data releases of the LAS [Lodieu et al. 2007b; Pinfield et al. 2008, Data Releases 1 and 2 (DR1 and DR2), respectively] to include all candidates drawn from DR4 (which incorporates DR1–DR3) which took place on 2008 July 1. Follow-up to confirm spectral types for this sample is now essentially complete for J≤ 19.0 and we report here the discovery of 47 new T dwarfs, including one T8+ dwarf and a number of spectrally peculiar objects, and use this sample to place improved constraints on the form of the field substellar mass spectrum.

2 SOURCE IDENTIFICATION

Candidate T dwarfs were selected from UKIDSS LAS DR3 and DR4 following a method similar to that described in Lodieu et al. (2007b), Pinfield et al. (2008) and Burningham et al. (2008, 2009). The UKIDSS project is defined in Lawrence et al. (2007), and uses the UKIRT Wide Field Camera (WFCAM; Casali et al. 2007). The photometric system is described in Hewett et al. (2006). The pipeline processing and science archive are described by Irwin et al. (2004) and Hambly et al. (2008).

Initial candidate selection is based on the presence of blue J−H and J−K colours (≤0.1) and red z′−J (>3.0, or non-detection in SDSS DR7). Since the LAS has nominal 5σ limits of 18.8 and 18.4 in the H and K bands, respectively (cf. 20.5 and 20.0 for Y and J), we ensure that we are complete to a faint cut-off at J= 19.3 by further selecting sources that are detected in YJ, but are undetected in HK, and also have z′−J > 3.0, or are undetected in SDSS DR7 (we only consider sources that lie within the SDSS footprint). In the case of the sources that are undetected in H and K, we also impose a Y−J > 0.5 requirement, to help eliminate contamination from earlier type M dwarfs near the faint limit of our selection.

To monitor for the existence of T dwarfs with Y−J < 0.5, we do not impose a Y−J colour cut on our YJH-selected candidates. However, following the cross-match against SDSS no candidates with Y−J < 0.5 remain in the sample (see Fig. 1). We thus infer that the YJ criterion for H and K non-detections should not exclude any T dwarfs.

Figure 1

DR4 T dwarf candidates plotted on a YJH colour–colour plot. YJH-selected candidates are shown as diamonds, whilst objects that are undetected in H in UKIDSS DR4 are shown as upper limits in JH. The photometry is that from the UKIDSS survey.

Open in new tabDownload slide

Candidates were then followed up photometrically to remove contamination and determine accurate magnitudes across the full zYJHK range where possible. In 700 deg² of DR3 and DR4 LAS sky [excluding 280 deg² of DR1 and DR2 sky already searched by Lodieu et al. (2007b) and Pinfield et al. (2008)], our initial SDSS + UKIDSS based selection identified 83 candidates with YJH detections (black diamonds in Fig. 1) and 80 candidates with YJ-only detections (indicated with arrows in Fig. 1), with J < 19.3.

Our follow-up is most complete for targets with J < 19.0, of which there were 55 YJH candidates and 17 YJ-only candidates. Of the 55 YJH candidates, photometric follow-up revealed three to be likely solar system objects (SSOs) which had apparently moved since the original survey were taken, and five were found to be objects with J−H > 0.1 that had been scattered into our selection by either photometric error or intrinsic variability (YJ and JH data were often taken at different epochs). Of the remaining 47 targets, 46 have been confirmed as T dwarfs and one still requires spectroscopic confirmation, 38 are presented here and eight have been previously published elsewhere (Burningham et al. 2008; Delorme et al. 2008; Pinfield et al. 2008; Burningham et al. 2009). Of the 17 YJ-only candidates with J < 19.0, four were found to be likely SSOs, seven were found to be objects with J−H > 0.1 but whose H-band magnitude was beyond the H-band limit of the LAS. Five have been confirmed as T dwarfs, and one target still requires photometric follow-up.

Although the follow-up of the J > 19.0 portion of our candidate list is less complete, it is predictably dominated by contamination caused by photometric scatter. Of the 25 YJH candidates with J > 19.0 that have been followed up photometrically, 22 have revised J−H > 0.1. For the YJ-only selection we find that of the 24 sources with J < 19.0 that have been followed up, 18 have J−H > 0.1, whilst two are likely SSOs.

The photometric follow-up required significant investment of observing time, dominated by deep H-band imaging of objects fainter than J= 19.0, which required 1-h integrations to achieve 10 per cent photometry on 4-m class facilities, while brighter candidates typically required 20–30 min integrations for similar results. All objects that passed the photometric follow-up stage were spectroscopically confirmed as T dwarfs.

2.1 Near-infrared photometry

Near-infrared follow-up photometry was obtained using the UKIRT Fast Track Imager (UFTI; Roche et al. 2003) mounted on UKIRT, and the Long-slit Infrared Imaging Spectrograph (LIRIS; Manchado et al. 1998) mounted on the William Herschel Telescope on La Palma, across a number of observing runs spanning 2007–2009. The final image mosaics were produced using sets of jittered images, with individual exposure times, jitter patterns and number of repeats given in Appendix A along with the dates of the observations. The data were dark subtracted, flat-field corrected, sky subtracted and mosaicked using ORAC-DR for the UFTI data and LIRIS-DR for the LIRIS data.

Data were obtained in a variety of observing conditions across the runs from photometric conditions with stable, subarcsecond seeing to thin cirrus with variable seeing. Under the photometric conditions photometric calibration was achieved using a standard star observed at similar airmass. Observations taken under non-photometric conditions were calibrated against 2MASS stars within the field of view.

The resulting photometry for the new T dwarfs discussed here is presented on the MKO system in Table 1. Where no follow-up photometry was obtained (mainly in the Y and K bands) the survey photometry is supplied instead. Data obtained on the LIRIS system were converted to the MKO system using transforms given in Pinfield et al. (2008) and Burningham et al. (2009). Appendix A lists the instruments used and dates for all sources described here, along with details of the conditions.

2.2 Optical photometry

We obtained optical z-band photometry using the European Southern Observatory (ESO) Multi-Mode Instrument (EMMI) mounted on the New Technology Telescope (NTT) at La Silla, Chile under programme 080.C-0090 and also using the ESO Faint Object Spectrograph and Camera (EFOSC2), also on the NTT, under programmes 081.C-0552 and 082.C-0399. The details of the observations for each target are given in Appendix A. The data were reduced using standard iraf packages, and then multiple images of the same target were aligned and stacked to increase signal-to-noise ratio. The zero-points were determined by using SDSS stars as secondary calibrators. The uncertainty we quote for our z-band photometry incorporates a scatter of ∼±0.05 in the determined zero-points.

For the EMMI follow-up, which used a Bessel z-band filter (ESO Z#611), we used the transform given by Warren et al. (2007) to calculate z_EMMI for the fiducial SDSS stars. The resulting EMMI photometry was then transformed to the Sloan z′(AB) system (see Warren et al. 2007).

For the EFOSC2 observations, for which a Gunn z-band filter (ESO Z#623) was used, we used the transform given in Burningham et al. (2009) to calculate z_EFOSC2 for the secondary calibrators. To place the resulting z_EFOSC2(AB) photometry on the Sloan z′(AB) system, we synthesized photometry for the T dwarf spectral standards (Burgasser et al. 2006), finding that across the T3–T8 range z_EFOSC2(AB) −z′(AB) =−0.19 ± 0.02, which is similar to the offset calculated by Warren et al. (2007) for z_EMMI(AB) −z′(AB) =−0.2.

The resulting photometry, on the z′(AB) system, is given in Table 1.

3 SPECTROSCOPIC CONFIRMATION

3.1 Observations and data reduction

Spectroscopic confirmation of those candidates that survived the photometric follow-up programme was achieved using the Near-InfraRed Imager and Spectrometer (NIRI; Hodapp et al. 2003) on the Gemini North Telescope 1 and the InfraRed Camera and Spectrograph (IRCS; Kobayashi et al. 2000) on the Subaru telescope, both on Mauna Kea, Hawaii. All observations were made up of a set of subexposures in an ABBA jitter pattern to facilitate effective background subtraction, with a slit width of 1 arcsec. The length of the AB jitter was 10 arcsec. The details of individual observations are summarized in Appendix A. The NIRI observations were reduced using standard iraf Gemini packages. The Subaru IRCS JH spectrum was also extracted using standard iraf packages. The AB pairs were subtracted using generic iraf tools, and median stacked.

Comparison argon arc frames were used to obtain dispersion solutions, which were then applied to the pixel coordinates in the dispersion direction on the images. The resulting wavelength-calibrated subtracted pairs had a low level of residual sky emission removed by fitting and subtracting this emission with a set of polynomial functions fit to each pixel row perpendicular to the dispersion direction, and considering pixel data on either side of the target spectrum only. The spectra were then extracted using a linear aperture, and cosmic rays and bad pixels removed using a sigma-clipping algorithm.

Telluric correction was achieved by dividing each extracted target spectrum by that of an early A or F type star observed just before or after the target and at a similar airmass. Prior to division, hydrogen lines were removed from the standard star spectrum by interpolating the stellar continuum. Relative flux calibration was then achieved by multiplying through by a blackbody spectrum of the appropriate T_eff. Data obtained for the same spectral regions on different nights were co-added after relative flux calibration, each weighted by their exposure time.

Initial, short, J- or JH-band spectra were used to establish an object's status as a T dwarf. In cases where the initial spectrum was suggestive of a late type, or the target had unusual colours for its type, deeper J-band and H- and K-band spectra were also obtained, and joined together using the measured near-infrared photometry to place the spectra on an absolute flux scale. In the case of Subaru/IRCS spectra, the red end of the spectra taken with the JH grism overlaps with blueward limit of the spectra taken with the HK grism. We were able to use the overlap region in the H band to bring the spectra on to a common flux scale for joining. Complete details of the spectroscopic observations obtained for each of the T dwarfs presented here are given in Appendix A. The resulting spectra are shown in Fig. 2.

Figure 2

Spectra of the 47 T dwarfs presented here. Each spectrum is normalized at 1.27 ± 0.005μm and offset for clarity. The spectra have been rebinned by a factor of 3 to maximize signal-to-noise ratio whilst not sacrificing resolution.

Open in new tabDownload slide

3.2 Spectral types

Spectral types have been determined through by-eye comparison to spectral type templates, and by measuring spectral type indices for the T sequence laid out in Burgasser et al. (2006) and extended by Burningham et al. (2008) for the latest types. The indices used here are summarized in Table 2. The final adopted type is arrived at by consideration of the determined indices and the best template match. The template comparison is generally given as much weight as the combined result of the indices. However, in cases where only two non-degenerate typing indices are available, and they disagree, the template comparison type is adopted. Fig. 3 shows the values of the computing indices plotted against adopted type for the sample presented here, along with previously identified T dwarfs.

Table 2

Open in new tab

The indices used for spectral typing the sample presented here. This follows the system for T0–T8 dwarfs described by (Burgasser et al. 2006) and extended to T9 by (Burningham et al. 2008) using the W_J index suggested by Warren et al. (2007).

Type .	H2O-J .	CH4-J .	WJ .	H2O-H .	CH4-H .	CH4-K .
T0	–	0.73–0.78	–	0.60–0.66	0.97–1.00	0.75–0.85
T1	>0.55	0.67–0.73	–	0.53–0.60	0.92–0.97	0.63–0.75
T2	0.45–0.55	0.58–0.67	–	0.46–0.53	0.80–0.92	0.55–0.63
T3	0.38–0.45	0.52–0.58	–	0.43–0.46	0.60–0.80	0.35–0.55
T4	0.32–0.38	0.45–0.52	–	0.37–0.43	0.48–0.60	0.24–0.35
T5	0.18–0.32	0.36–0.45	–	0.32–0.37	0.36–0.48	0.18–0.24
T6	0.13–0.18	0.28–0.36	–	0.26–0.32	0.25–0.36	0.13–0.18
T7	0.07–0.13	0.21–0.28	0.35–0.40	0.20–0.26	0.15–0.25	<0.13
T8	<0.07	<0.21	0.28–0.35	0.14–0.20	<0.15	–
T9	–	–	<0.28	<0.14	–	–

Type .	H2O-J .	CH4-J .	WJ .	H2O-H .	CH4-H .	CH4-K .
T0	–	0.73–0.78	–	0.60–0.66	0.97–1.00	0.75–0.85
T1	>0.55	0.67–0.73	–	0.53–0.60	0.92–0.97	0.63–0.75
T2	0.45–0.55	0.58–0.67	–	0.46–0.53	0.80–0.92	0.55–0.63
T3	0.38–0.45	0.52–0.58	–	0.43–0.46	0.60–0.80	0.35–0.55
T4	0.32–0.38	0.45–0.52	–	0.37–0.43	0.48–0.60	0.24–0.35
T5	0.18–0.32	0.36–0.45	–	0.32–0.37	0.36–0.48	0.18–0.24
T6	0.13–0.18	0.28–0.36	–	0.26–0.32	0.25–0.36	0.13–0.18
T7	0.07–0.13	0.21–0.28	0.35–0.40	0.20–0.26	0.15–0.25	<0.13
T8	<0.07	<0.21	0.28–0.35	0.14–0.20	<0.15	–
T9	–	–	<0.28	<0.14	–	–

Table 2

Open in new tab

The indices used for spectral typing the sample presented here. This follows the system for T0–T8 dwarfs described by (Burgasser et al. 2006) and extended to T9 by (Burningham et al. 2008) using the W_J index suggested by Warren et al. (2007).

Type .	H2O-J .	CH4-J .	WJ .	H2O-H .	CH4-H .	CH4-K .
T0	–	0.73–0.78	–	0.60–0.66	0.97–1.00	0.75–0.85
T1	>0.55	0.67–0.73	–	0.53–0.60	0.92–0.97	0.63–0.75
T2	0.45–0.55	0.58–0.67	–	0.46–0.53	0.80–0.92	0.55–0.63
T3	0.38–0.45	0.52–0.58	–	0.43–0.46	0.60–0.80	0.35–0.55
T4	0.32–0.38	0.45–0.52	–	0.37–0.43	0.48–0.60	0.24–0.35
T5	0.18–0.32	0.36–0.45	–	0.32–0.37	0.36–0.48	0.18–0.24
T6	0.13–0.18	0.28–0.36	–	0.26–0.32	0.25–0.36	0.13–0.18
T7	0.07–0.13	0.21–0.28	0.35–0.40	0.20–0.26	0.15–0.25	<0.13
T8	<0.07	<0.21	0.28–0.35	0.14–0.20	<0.15	–
T9	–	–	<0.28	<0.14	–	–

Type .	H2O-J .	CH4-J .	WJ .	H2O-H .	CH4-H .	CH4-K .
T0	–	0.73–0.78	–	0.60–0.66	0.97–1.00	0.75–0.85
T1	>0.55	0.67–0.73	–	0.53–0.60	0.92–0.97	0.63–0.75
T2	0.45–0.55	0.58–0.67	–	0.46–0.53	0.80–0.92	0.55–0.63
T3	0.38–0.45	0.52–0.58	–	0.43–0.46	0.60–0.80	0.35–0.55
T4	0.32–0.38	0.45–0.52	–	0.37–0.43	0.48–0.60	0.24–0.35
T5	0.18–0.32	0.36–0.45	–	0.32–0.37	0.36–0.48	0.18–0.24
T6	0.13–0.18	0.28–0.36	–	0.26–0.32	0.25–0.36	0.13–0.18
T7	0.07–0.13	0.21–0.28	0.35–0.40	0.20–0.26	0.15–0.25	<0.13
T8	<0.07	<0.21	0.28–0.35	0.14–0.20	<0.15	–
T9	–	–	<0.28	<0.14	–	–

Figure 3

Adopted spectral types plotted against computed index values (see Table 2) for the sample presented here, and previously published T dwarfs from Burgasser et al. (2006), Warren et al. (2007), Lodieu et al. (2007b), Delorme et al. (2008), Burningham et al. (2008), Pinfield et al. (2008), Burningham et al. (2009) and Burningham et al. (2010). Previously published T dwarfs are indicated with ‘+’ symbols. Types derived from NIRI spectra are indicated by circles and from IRCS data by squares. Blue symbols represent CH₄-J-early peculiar objects, whilst red symbols indicate H₂O-H-early peculiarity. Colour version of figure is available in on-line version.

Open in new tabDownload slide

The results of this process are given in Table 3. In cases where spectroscopy is available across the entire JHK range, the uncertainties in the types are ±0.5 subtypes. In the majority of cases, however, we only have coverage in J or JH, and in these cases the precision drops to ±1 subtype.

Two objects in our selection have recently been independently identified as T dwarfs by Goldman et al. (2010): ULAS J1149−0143 and ULAS J1153−0147. Using methane imaging they estimate spectral types of T5 ± 1.5 and T6.5 ± 1 for these objects, respectively. These estimates agree well with our derived spectral types of T5 and T6 (±1).

3.3 Peculiar classifications

We have classified six objects as peculiar in Table 3, as indicated by the postscript ‘p’ on their assigned type. There are essentially two routes to classification as peculiar: (i) a mismatch of two or more subtypes between the results of the different spectral typing ratios; (ii) other specific anomalies in the spectrum as compared to the template spectrum for the adopted type. Of the seven objects classified as peculiar in Table 3, all exhibit a spectral type mismatch.

The mismatch of spectral types implied by different indices has been seen before. In Pinfield et al. (2008), the T6.5p dwarf ULAS J1150+0949 displayed a T3-like H₂O-H index, whilst displaying later types in all other indices. The T8p dwarf ULAS J1017+0118 (Burningham et al. 2008) displays a similar mismatch, with indices consistent with T8 classification in the J band, but T6-like indices in the H and K bands. Gl 229B displays T5-like indices in H₂O-H, and CH₄-K, a T6-like H₂O-J and T7-like CH₄-J and CH₄-H (Burgasser et al. 2006).

In the sample presented here, the peculiar objects can be broadly separated into two classes. For the purposes of this discussion we define these as follows.

H₂O-H-early: H₂O-H index implies an earlier type than that suggested by the H₂O-J index by at least two subtypes (see Fig. 4).

Figure 4

The spectrum of the H₂O-H-early T6p dwarf ULAS J0853+0006, plotted with spectra for the T6 and T2 spectral templates. The error spectrum for ULAS J0853+0006 is shown as a black line offset below the zero flux line.

Open in new tabDownload slide

CH₄-J-early: CH₄-J index implies an earlier type than that suggested by the H₂0-J index by at least two subtypes (see Fig. 5).

Figure 5

The spectrum of CH₄-J-early peculiar T7p dwarf ULAS J1034−0015 compared to the T7 template 2MASS J0727+1710. The error spectrum for ULAS J1034−0015 is shown as a black line offset below the zero flux line.

Open in new tabDownload slide

Figs 6 and 7 show comparisons of the indices that define these peculiarities. In both cases the peculiar objects are apparent as outliers. It is also apparent that not all outliers have been classified as peculiar. This is due to the fact that the peculiarities have been defined with respect to the spectral index bins for each subtype, the size of which is not the same for all subtypes and differs between the ratios themselves. As a result an object can appear as an outlier in Figs 6 and 7, although its index values might be consistent to within a subtype. The definitions of the peculiarities described above may, in the future, require refinement but for now will serve as reasonable starting points for highlighting the objects that exhibit the most obvious peculiar features.

Figure 6

A comparison of the H₂O-J and H₂O-H spectral typing indices for the T dwarfs presented here and those published elsewhere. Symbols are as for Fig. 3.

Open in new tabDownload slide

Figure 7

A comparison of the H₂O-J and CH₄-J spectral typing indices for the T dwarfs presented here and those published elsewhere. Symbols are as for Fig. 3.

Open in new tabDownload slide

In addition to the two objects already published, we have identified two T6p objects and two T5p objects which display H₂O-H-early peculiarity. One additional peculiar object has been identified for which the CH₄-J index disagrees with the H₂O-H index, and as such does not fall within the two classes of peculiarity defined above.

For objects near the L–T transition, the combination of an earlier spectral type in the H band than that seen in the J band can be caused by unresolved binarity. In these cases the secondary T dwarf dominates in the J band due to the J-band brightening observed through the early T sequence (T1–T5) (e.g. Dahn et al. 2002; Tinney, Burgasser & Kirkpatrick 2003; Vrba et al. 2004), whilst the primary contributes more flux in the H band (e.g. Looper et al. 2008). This scenario is a potential explanation for the unusual spectral properties of ULAS J0943+0858 and ULAS J1319+1209, which both appear as T5 in the J band, but T3 in the H band.

To assess if this is the origin of the H₂O-H-early peculiarity seen in our sample, we have simulated various combinations of L8+T binaries (since the local minimum is L8 for M_J in the L–T transition providing the maximum boost to the secondary) and have found that a combination of a T6 and an L8 dwarf can produce a similar combination of spectral type indices as seen in ULAS J0943+0858. However, it requires the T6 to be at least 0.75 mag brighter than the L8 dwarf, and as such it would be a significant outlier on the M_J versus spectral type produced by Dahn et al. (2002), Tinney et al. (2003) and Vrba et al. (2004). Given the difficulty in producing a T5p through unresolved binarity, it comes as little surprise that we are also unable to simulate the T6p dwarfs by this mechanism. Since binarity alone is unlikely to cause H₂O-H-early peculiarity, we speculate that this morphology may instead represent an as-yet unidentified tracer of composition and/or surface gravity.

Of the four CH₄-J-early peculiar objects identified here, the 1.25 μm K i doublet is apparently absent from two (ULAS J1231+0912 and ULAS J1034−0015); however, the quality of the spectra are too poor to rule out its presence. Given such a small sample of objects displaying peculiarity of this type, interpretation of its significance is problematic. Additionally, the low signal-to-noise ratio of these spectra raises the possibility that the peculiarity is merely a product of poor-quality spectra.

The steep gradient and low flux in the spectrum in the region for the denominator of the CH₄-J index suggests that this peculiarity may in some cases be spurious. For example, the initial discovery spectrum of ULAS J1233+1219 displayed apparent CH₄-J-early peculiarity. However, the subsequent deeper follow-up spectrum resulted in spectral typing indices that were in agreement (although the red side of the J-band peak still appears somewhat earlier in type than the indices would otherwise imply). Although these issues are of concern, the presence of CH₄-J-early peculiarity in higher signal-to-noise ratio spectra (e.g. Burningham et al. 2010) argues for its inclusion with the types assigned in this paper.

Higher signal-to-noise ratio spectra of the peculiar objects identified here will establish the significance of the peculiarities discussed above. Determination of their kinematic properties should also be a priority, allowing for identification of any wide common proper-motion companions capable of providing fiducial constraints on metallicity and age.

3.4 ULAS J1302+1308 – a new T8.5 dwarf

Our newly expanded sample of LAS T dwarfs includes a new T8.5 dwarf, ULAS J1302+1308. This objects brings the number of published T8+ objects to six (Warren et al. 2007; Burningham et al. 2008; Delorme et al. 2008; Burningham et al. 2009). With a growing sample of all late-type objects we thus return to the question of the presence of ammonia absorption at 1.58 μm, first suggested by Delorme et al. (2008). In Fig. 8 we plot the suggested ‘NH₃′-H index of Delorme et al. (2008) against spectral type [on the Burgasser et al. (2006) and Burningham et al. (2008) system]. Whilst this supports the assertion that the new index is effective at distinguishing the latest type objects, it does not strongly suggest the introduction of a new opacity source at later types. The trend is very much a continuation (with a similar degree of scatter) of that seen at earlier types, where opacity in this region is largely attributed to water vapour.

Figure 8

The ‘NH₃’-H index for T6–T9 dwarfs classified on the system of Burgasser et al. (2006), extended to late types by Burningham et al. (2009). The plotted values incorporate objects described in this paper and previously published objects (Burgasser et al. 2006; Looper, Kirkpatrick & Burgasser 2007; Warren et al. 2007; Burningham et al. 2008; Delorme et al. 2008; Burningham et al. 2009). Index values for the latter have been calculated from the objects’ spectra supplied by the authors. Uncertainties in the index are smaller than the symbol size, whilst uncertainties in the spectral type are typically ±0.5 subtypes. The open diamond symbols indicate objects classified as peculiar.

Open in new tabDownload slide

4 THE NEAR-INFRARED COLOURS OF UKIDSS T DWARFS

In addition to the 47 new T dwarfs presented here, we will consider all T dwarfs thus far published that lie within the LAS DR4 footprint (Kendall et al. 2007; Lodieu et al. 2007b; Warren et al. 2007; Burningham et al. 2008; Chiu et al. 2008; Delorme et al. 2008; Pinfield et al. 2008; Burningham et al. 2009), giving a total of 80 objects with spectral types from T0 to T9. Fig. 9 shows various far-red and near-infrared colours plotted as a function of spectral type.

Figure 9

Spectral type versus colour plots for the T dwarfs presented in this paper and all other published T dwarfs from the UKIDSS LAS. T dwarfs classified as peculiar in Section 3 are plotted in the red and blue indicating H₂O-H-early and CH₄-J-early peculiarities, respectively.

Open in new tabDownload slide

Most striking is the wide scatter present in all colours for late-T dwarfs. This indicates that within each spectral type there is significant diversity of properties such as T_eff, gravity and metallicity. However, trends with increasing spectral type are apparent in some colours, suggesting changes that are driven principally by decreasing T_eff. It can be seen that the trends in Y−J and J−H already noted by Leggett et al. (2010) are reflected in our expanded sample. Also confirmed is the wide scatter in H−K and J−K. A number of objects are noteworthy on these latter two plots as lying blueward of the bulk population. These objects may be members of a low-metallicity population and we defer detailed examination of these objects’ spectra and kinematics to a future work (Murray et al., in preparation).

There appears to be a weak continuation of the trend of reddening z′−J with type seen in earlier type objects (e.g. Chiu et al. 2006), but thus far undetected for the latest spectral types (Pinfield et al. 2008), presumably due to its broad scatter and the previously small sample size in this range. Most interesting, however, is the relatively strong trend in z′−Y for ≥T6 dwarfs, coupled with decreasing Y−J in the same regime. The latter has been noted elsewhere (e.g. Leggett et al. 2010); however, it was not previously clear if this was caused by depression of the J-band peak, or brightening in the Y band. The combined trends seen in Fig. 9 suggest the brightening of the Y band is responsible. This is most likely due to the broad K i absorption at 0.77 μm weakening as K i condenses into KCl (Lodders 1999).

We have also indicated the objects classified as peculiar in Section 3. Although some of the peculiar objects are clear outliers from the bulk population, the majority appear to have fairly typical colours for their types. There is also no apparent distinction between the two flavours of peculiarity discussed in Section 3. Unfortunately, we only have K-band photometry for one of the CH₄-J-early peculiar objects. This object has extremely blue H−K, consistent with the low-metallicity/high-gravity interpretation of its spectral morphology. Clearly more complete photometry, and better spectral coverage, for these unusual objects is required before sound physical interpretation of their properties will be possible.

5 CONSTRAINING THE SUBSTELLAR MASS FUNCTION

In Pinfield et al. (2008) comparison of the DR2 sample of ≥T4 dwarfs with the simulations of Deacon & Hambly (2006) allowed weak constraints to be placed on the substellar field IMF, favouring α≤ 0 for an IMF of the form Ψ (M) ∝M^−α. The increased volume probed by our DR4 sample allows us to now examine for the first time the constraints that may be drawn from a ≥T6 sample.

5.1 The sample

The sample of T dwarfs considered below represents all LAS T dwarfs that have been spectroscopically confirmed from 980 deg² of DR4 sky. All candidates with J < 19.0 and confirmed J−H < 0.1 that arose in our selection have been followed up, with the exception of one source. The source is a target from our so-called YJ-only search with J= 18.8. Contamination at this magnitude for the YJ-only search is typically worse than 60 per cent, so we do not include this object in our final sample. The total number of T dwarfs identified to date in LAS DR4 sky with J−H < 0.1 is 80 (Burgasser et al. 2004; Lodieu et al. 2007b; Burningham et al. 2008; Delorme et al. 2008; Pinfield et al. 2008; Burningham et al. 2009). Of these 31 have J= 18.5–19.0 and 42 have J < 18.5. If our sample were essentially complete to J= 19.0, we would expect our sample size to double with an increased depth of 0.5 mag (since this doubles the survey volume). It thus seems likely that our sample is not complete for late-T dwarfs to J= 19.0, as had been claimed by Pinfield et al. (2008), and we instead restrict our analysis to a brighter sample.

By selecting T dwarfs with J≤ 18.8 we find that 25 have 18.3 ≤ 18.8 and 36 with J < 18.3. Assuming Poisson noise, these numbers are roughly consistent with a similar level of completeness in the two bins, although it is likely that we are somewhat incomplete in the fainter bin. The completeness of our T dwarf sample is set by the Y-band completeness which is the fainter of our required detections in Y and J. Completeness as a function of signal-to-noise ratio has been modelled through comparison of fields that have been observed in both UKIDSS Deep extragalactic Survey (DXS) and the LAS. This process suggests that at Y= 20.3, which corresponds to our J= 18.8 limit for T dwarfs with Y−J= 1.5, 68 per cent of sources are detected in the LAS. Only 4 per cent of our T dwarfs have Y−J > 1.5, and 60 per cent have Y−J < 1.3. For this latter group, the DXS–LAS comparison suggests that we are 85 per cent complete at J= 18.8. For J < 18.3 we can expect to be 97 per cent complete.

To assess if the depth achieved in the DXS overlap fields is typical, and how much scatter there is across LAS DR4, we have also determined the distribution of mean Y, J and H magnitudes for the LAS fields. Approximately 6 per cent of J-band fields and 9 per cent of Y-band fields have mean magnitudes that are more than 0.3 mag brighter than the mean value for the entire DR4 LAS. Additionally, similar ratios of source counts across the four YJHK bands are seen for all fields. We thus conclude that the variation in depth and completeness across the LAS is relatively minor, and that our sample should be complete at the 85 per cent level at its faintest limit.

Our sample has been selected in an identical manner to that of Pinfield et al. (2008), and we follow a similar method for accounting for sources of incompleteness and bias. The selection method is most sensitive to late-T dwarfs, for which J−H < 0.1. Through consideration of the SDSS iz-selected T dwarfs of Chiu et al. (2006) and Pinfield et al. (2008) demonstrated that we could expect to exclude <10 per cent of ≥T4 dwarfs with this selection. Fig. 9 indicates that this proportion is likely to be much smaller for spectral types ≥T6 (see below). The UKIDSS LAS sample has now grown sufficiently that we are able to select just ≥T6 dwarfs and still have a workable sample size. As such, we now diverge from the treatment of Pinfield et al. (2008) and focus on a sample of ≥T6 dwarfs with J≤ 18.8, of which we have identified 25 to date in the LAS.

Since only one ≥T6 dwarf in our sample has J−H > −0.2, we proceed on the basis that all such objects have J−H < 0.1, and in absence of photometric scatter would be included in our colour selection. However, we have estimated that we could expect one ≥T6 dwarf to be scattered out of our colour selection by photometric measurement errors 2 leading to an estimated sample of 26 ≥ T6 dwarfs.

Since we have obtained spectroscopy for the entire sample that we are considering, we do not need to correct for contaminants that have been scattered into our initial photometric selection, as would be the case for a purely colour-selected sample. However, uncertainties in spectral typing will lead to objects being both scattered out of our sample, and scattered into it. The larger number of detected T5.5 objects compared to T6 objects suggests that with 0.5 subtype uncertainties there should be a small net result of contaminants scattered into our sample. However, we estimate that fewer than one contaminant will be introduced to our sample and we thus neglect this effect from further consideration.

An additional source of incompleteness in our sample arises from the manner in which our selection relies on cross-matching SDSS with UKIDSS. Our selection requires sources to have z′−J > 3.0, or to be undetected in SDSS, a status which is established by searching for optical counterparts within 2 arcsec of the UKIDSS sources. The raises the possibility that bona fide candidates which are, in reality, non-detections in SDSS will be rejected from our selection if an unrelated optical source is close enough to be misidentified as an optical counterpart. This issue was examined in Pinfield et al. (2008) for the case of UKIDSS–SDSS cross-matching, and a correction of +3 per cent was found to account for this source of incompleteness.

In addition to this effect, the brightest stars effectively mask the sky in both the LAS and SDSS, hiding potential candidates and preventing effective assessment of their optical properties in SDSS. We have estimated that stars with J < 12.0 mask a disc on the sky with a typical radius ∼10 arcsec. There are ∼10⁵ such stars in UKIDSS LAS DR4, masking less than 1 per cent of the sky from our search method. We thus do not correct for this source of incompleteness.

To allow comparison of our sample to simulations we now remove objects that we know are binary companions to higher mass objects. In our current sample of ≥T6 objects, there is just the T8.5 dwarf Wolf 940B so far identified as a wide companion to a main sequence star. Given that approximately 4 per cent of published T dwarfs are wide companions to higher mass stars (e.g. http://www.DwarfArchives.org), it is likely that removing this object from our sample is consistent with the required correction.

A further correction is required to account for the inclusion of unresolved binary systems in our magnitude-limited sample, the components of which would fall beyond our J < 18.8 cut, were they single objects, but which are included in our magnitude-limited sample because their combined luminosity makes them visible at greater distances. If we define our observed binary fraction as (following Burgasser et al. 2003)

1

where N_B and N_m are the number of binaries and the total number of sources in our magnitude-limited sample, respectively; BF is the underlying ‘true’ binary fraction and γ is the fractional increase in volume from which binaries are selected. The fraction of binaries that should be excluded from the sample as they lie beyond the distance suggested by our J= 18.8 mag limited is

2

where N_D is the number of binaries that fall within the magnitude limit due to their overluminosity. The fraction of sources that must be excluded from our magnitude-limited sample is thus found as

3

A number of estimates are available for the binary fraction of field and young cluster brown dwarfs (e.g. Burgasser et al. 2003; Pinfield et al. 2003; Maxted & Jeffries 2005; Allen 2007; Lodieu et al. 2007a) ranging from ∼10 to 50 per cent. For our upper limit we take results of the study by Maxted & Jeffries (2005), who used Monte Carlo simulation techniques and the results of published radial velocity surveys to estimate a binary fraction of 32–45 per cent. For our lower limit we use the result of Burgasser et al. (2003), who estimated a binary fraction of 5–24 per cent from high-resolution imaging of field T dwarfs.

Burgasser et al. (2003) derived values for γ based on different assumptions about flux ratio distribution. We will take their values for two extreme cases: that of a flat flux ratio distribution (γ= 1.9) and the case where all binaries consist of equal flux components (⁠⁠). We thus find that we should exclude 3–45 per cent of the sources in our sample to account for the presence of unresolved binarity, resulting in a binary corrected sample of 14–25 ≥T6 dwarfs.

In Pinfield et al. (2008) it was decided that only counting the primary, when most T dwarf binaries appear to be equal-mass binary systems, would incorrectly exclude T dwarf secondaries. The correction that was applied in that case not only excluded binary systems that lay beyond the magnitude limit, but also counted the secondaries of binary systems which lay within the limit. The range of corrections that resulted lie within the range of corrections derived above, so we do not include this consideration in our treatment.

Finally, we apply a correction factor to account for the Malmquist bias. Pinfield et al. (2008) derived a correction by which the sample of T dwarfs was reduced by 12–16 per cent, which results in final estimates of 11–22 (±5) ≥T6 dwarfs with J < 18.8 in the 980 deg² of DR4 sky.

To allow a convenient comparison with results from 2MASS–SDSS cross-matching (Metchev et al. 2008), the CFBDS (Reylé et al. 2009) and simulations such as those by Burgasser (2004) we now estimate the space density of T6–T9 dwarfs. Table 4 shows the calculated space densities for spectral types between T6 and T9, applying the same corrections to the bins for each subtype as we previously applied to the whole sample. The smaller and larger estimates reflect extremes of the range of possible binary fractions, corrected for Malmquist bias.

Summing the estimates for the T6–T8 range suggests a lower space density limit of 1.3 ± 0.6 × 10⁻³ pc⁻³, and an upper limit of 2.5 ± 1.2 × 10⁻³ pc⁻³. The uncertainties reflect Poisson noise in the sample, the range of spectral types in each bin and the dispersion of the M_J–spectral type relation of Liu et al. (2006). Our space density estimate is lower, but formally consistent with, the value found by Metchev et al. (2008) of 4.7^+3.1_−2.8× 10⁻³ pc⁻³.

We provide two estimates for the space density of T9 dwarfs. The first was estimated using a value of M_J= 17.4–18.1 (by extending the polynomials of Liu et al. 2006, to later types for the T9–T9.5 range), and so is on the same ‘system’ as the earlier types. This results in a space density for the T9 dwarfs of 3.1–6.1 × 10⁻³ pc⁻³. However, the first parallax measurements for T8+ dwarfs suggest M_J= 17.6–18.2 for these objects (Burningham et al. 2009; Smart, private communication). Using these values results in a higher space density of 3.9–7.6 × 10⁻³ pc⁻³.

5.2 The simulations

To place constraints on the functional form of the IMF, we now compare our sample to predictions from Monte Carlo simulations of the field brown dwarf population. The simulations we will use are based on those of Deacon & Hambly (2006) and we refer the reader to that work for a detailed description. However, the simulations have been updated in the following ways.

Previous work, such as Burgasser (2004), Deacon & Hambly (2006) and Pinfield et al. (2008) have used the individual object mass function for normalization. While this will produce an accurate representation of the distribution of objects in the stellar luminosity function and the number of objects lying within a survey volume, it does not take into account objects obscured by brighter close binary components and hence will not reflect the potential number of detections from a wide-field search such as this study. We instead use the system mass function normalization 0.0024 pc⁻³ (for objects in the 0.09–0.1 M_⊙ range) taken from Deacon, Nelemans & Hambly (2008) which is consistent with the system mass function derived by Chabrier (2005).

Each simulated object was drawn from age and mass probability distributions defined by the various underlying birthrates and mass spectra that were simulated. The mass spectra were of the form

4

And the expression for the birthrate was of the form

5

T_eff were determined using the COND evolutionary models of Baraffe et al. (2003). Conversion of these T_eff into spectral types however is currently problematic, given the limited number of late-type dwarfs with well-determined parallaxes and good mid-infrared spectral coverage. Furthermore, there is significant scatter in the spectral type-T_eff relation for the objects that do have well-determined properties (e.g. Golimowski et al. 2004; Vrba et al. 2004), presumably as a result of the effects of varying composition and surface gravity.

As such, we have used a semi-empirical method for converting the simulated objects’T_eff to spectral types. The relevant conversions are given in Table 4, and were determined by reference to Vrba et al. (2004), Golimowski et al. (2004) and more recent studies of T8+ dwarfs by Leggett et al. (2009) and Burningham et al. (2009).

Finally, the simulated population has absolute magnitudes assigned as a function of spectral type using the relations of Liu et al. (2006) for the K band. Colours for each simulated object have been drawn from distribution based on the observed colours of our sample, with an additional scatter of 0.15 mag to account for the fact that the colours of each spectral type do not display a purely Gaussian scatter.

The resulting predictions for the number of ≥T6 dwarfs with J−H < 0.1 and J≤ 18.8 identified in our search of UKIDSS LAS DR4 for different underlying mass functions and birthrates are given in Table 5. In the relatively near future, the advent of the Wide-field Infrared Survey Explorer (WISE; Wright 2008) and ongoing warm-Spitzer programmes will allow comparison of samples such as ours to simulations in a considerably more robust manner. The H−[4.7] colour available from the WISE data set is likely to be as useful a T_eff indicator as the H−[4.5] has been (e.g. Warren et al. 2007) in the Spitzer era, and these programmes will allow determination of relatively robust T_eff for our complete (and future) sample. Thus, the somewhat unsatisfactory conversion of the simulations into spectral type space will no longer be necessary.